Gumbel-Softmax is a continuous distribution that has the property that it can be smoothly annealed into a categorical distribution, and whose parameter gradients can be easily computed via the reparameterization trick.
Source: Categorical Reparameterization with Gumbel-SoftmaxPaper | Code | Results | Date | Stars |
---|
Task | Papers | Share |
---|---|---|
Clustering | 6 | 8.70% |
Image Classification | 4 | 5.80% |
Speech Recognition | 4 | 5.80% |
General Classification | 4 | 5.80% |
Reinforcement Learning (RL) | 3 | 4.35% |
Quantization | 3 | 4.35% |
Recommendation Systems | 3 | 4.35% |
Multi-agent Reinforcement Learning | 2 | 2.90% |
Time Series Analysis | 2 | 2.90% |
Component | Type |
|
---|---|---|
🤖 No Components Found | You can add them if they exist; e.g. Mask R-CNN uses RoIAlign |